1/* SPDX-License-Identifier: GPL-2.0 */
2
3#ifndef BTRFS_BLOCK_GROUP_H
4#define BTRFS_BLOCK_GROUP_H
5
6#include <linux/atomic.h>
7#include <linux/mutex.h>
8#include <linux/list.h>
9#include <linux/spinlock.h>
10#include <linux/refcount.h>
11#include <linux/wait.h>
12#include <linux/sizes.h>
13#include <linux/rwsem.h>
14#include <linux/rbtree.h>
15#include <uapi/linux/btrfs_tree.h>
16#include "free-space-cache.h"
17
18struct btrfs_chunk_map;
19struct btrfs_fs_info;
20struct btrfs_inode;
21struct btrfs_trans_handle;
22
23enum btrfs_disk_cache_state {
24	BTRFS_DC_WRITTEN,
25	BTRFS_DC_ERROR,
26	BTRFS_DC_CLEAR,
27	BTRFS_DC_SETUP,
28};
29
30enum btrfs_block_group_size_class {
31	/* Unset */
32	BTRFS_BG_SZ_NONE,
33	/* 0 < size <= 128K */
34	BTRFS_BG_SZ_SMALL,
35	/* 128K < size <= 8M */
36	BTRFS_BG_SZ_MEDIUM,
37	/* 8M < size < BG_LENGTH */
38	BTRFS_BG_SZ_LARGE,
39};
40
41/*
42 * This describes the state of the block_group for async discard.  This is due
43 * to the two pass nature of it where extent discarding is prioritized over
44 * bitmap discarding.  BTRFS_DISCARD_RESET_CURSOR is set when we are resetting
45 * between lists to prevent contention for discard state variables
46 * (eg. discard_cursor).
47 */
48enum btrfs_discard_state {
49	BTRFS_DISCARD_EXTENTS,
50	BTRFS_DISCARD_BITMAPS,
51	BTRFS_DISCARD_RESET_CURSOR,
52};
53
54/*
55 * Control flags for do_chunk_alloc's force field CHUNK_ALLOC_NO_FORCE means to
56 * only allocate a chunk if we really need one.
57 *
58 * CHUNK_ALLOC_LIMITED means to only try and allocate one if we have very few
59 * chunks already allocated.  This is used as part of the clustering code to
60 * help make sure we have a good pool of storage to cluster in, without filling
61 * the FS with empty chunks
62 *
63 * CHUNK_ALLOC_FORCE means it must try to allocate one
64 *
65 * CHUNK_ALLOC_FORCE_FOR_EXTENT like CHUNK_ALLOC_FORCE but called from
66 * find_free_extent() that also activaes the zone
67 */
68enum btrfs_chunk_alloc_enum {
69	CHUNK_ALLOC_NO_FORCE,
70	CHUNK_ALLOC_LIMITED,
71	CHUNK_ALLOC_FORCE,
72	CHUNK_ALLOC_FORCE_FOR_EXTENT,
73};
74
75/* Block group flags set at runtime */
76enum btrfs_block_group_flags {
77	BLOCK_GROUP_FLAG_IREF,
78	BLOCK_GROUP_FLAG_REMOVED,
79	BLOCK_GROUP_FLAG_TO_COPY,
80	BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
81	BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED,
82	BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
83	BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
84	/* Does the block group need to be added to the free space tree? */
85	BLOCK_GROUP_FLAG_NEEDS_FREE_SPACE,
86	/* Indicate that the block group is placed on a sequential zone */
87	BLOCK_GROUP_FLAG_SEQUENTIAL_ZONE,
88	/*
89	 * Indicate that block group is in the list of new block groups of a
90	 * transaction.
91	 */
92	BLOCK_GROUP_FLAG_NEW,
93};
94
95enum btrfs_caching_type {
96	BTRFS_CACHE_NO,
97	BTRFS_CACHE_STARTED,
98	BTRFS_CACHE_FINISHED,
99	BTRFS_CACHE_ERROR,
100};
101
102struct btrfs_caching_control {
103	struct list_head list;
104	struct mutex mutex;
105	wait_queue_head_t wait;
106	struct btrfs_work work;
107	struct btrfs_block_group *block_group;
108	/* Track progress of caching during allocation. */
109	atomic_t progress;
110	refcount_t count;
111};
112
113/* Once caching_thread() finds this much free space, it will wake up waiters. */
114#define CACHING_CTL_WAKE_UP SZ_2M
115
116struct btrfs_block_group {
117	struct btrfs_fs_info *fs_info;
118	struct inode *inode;
119	spinlock_t lock;
120	u64 start;
121	u64 length;
122	u64 pinned;
123	u64 reserved;
124	u64 used;
125	u64 delalloc_bytes;
126	u64 bytes_super;
127	u64 flags;
128	u64 cache_generation;
129	u64 global_root_id;
130
131	/*
132	 * The last committed used bytes of this block group, if the above @used
133	 * is still the same as @commit_used, we don't need to update block
134	 * group item of this block group.
135	 */
136	u64 commit_used;
137	/*
138	 * If the free space extent count exceeds this number, convert the block
139	 * group to bitmaps.
140	 */
141	u32 bitmap_high_thresh;
142
143	/*
144	 * If the free space extent count drops below this number, convert the
145	 * block group back to extents.
146	 */
147	u32 bitmap_low_thresh;
148
149	/*
150	 * It is just used for the delayed data space allocation because
151	 * only the data space allocation and the relative metadata update
152	 * can be done cross the transaction.
153	 */
154	struct rw_semaphore data_rwsem;
155
156	/* For raid56, this is a full stripe, without parity */
157	unsigned long full_stripe_len;
158	unsigned long runtime_flags;
159
160	unsigned int ro;
161
162	int disk_cache_state;
163
164	/* Cache tracking stuff */
165	int cached;
166	struct btrfs_caching_control *caching_ctl;
167
168	struct btrfs_space_info *space_info;
169
170	/* Free space cache stuff */
171	struct btrfs_free_space_ctl *free_space_ctl;
172
173	/* Block group cache stuff */
174	struct rb_node cache_node;
175
176	/* For block groups in the same raid type */
177	struct list_head list;
178
179	refcount_t refs;
180
181	/*
182	 * List of struct btrfs_free_clusters for this block group.
183	 * Today it will only have one thing on it, but that may change
184	 */
185	struct list_head cluster_list;
186
187	/*
188	 * Used for several lists:
189	 *
190	 * 1) struct btrfs_fs_info::unused_bgs
191	 * 2) struct btrfs_fs_info::reclaim_bgs
192	 * 3) struct btrfs_transaction::deleted_bgs
193	 * 4) struct btrfs_trans_handle::new_bgs
194	 */
195	struct list_head bg_list;
196
197	/* For read-only block groups */
198	struct list_head ro_list;
199
200	/*
201	 * When non-zero it means the block group's logical address and its
202	 * device extents can not be reused for future block group allocations
203	 * until the counter goes down to 0. This is to prevent them from being
204	 * reused while some task is still using the block group after it was
205	 * deleted - we want to make sure they can only be reused for new block
206	 * groups after that task is done with the deleted block group.
207	 */
208	atomic_t frozen;
209
210	/* For discard operations */
211	struct list_head discard_list;
212	int discard_index;
213	u64 discard_eligible_time;
214	u64 discard_cursor;
215	enum btrfs_discard_state discard_state;
216
217	/* For dirty block groups */
218	struct list_head dirty_list;
219	struct list_head io_list;
220
221	struct btrfs_io_ctl io_ctl;
222
223	/*
224	 * Incremented when doing extent allocations and holding a read lock
225	 * on the space_info's groups_sem semaphore.
226	 * Decremented when an ordered extent that represents an IO against this
227	 * block group's range is created (after it's added to its inode's
228	 * root's list of ordered extents) or immediately after the allocation
229	 * if it's a metadata extent or fallocate extent (for these cases we
230	 * don't create ordered extents).
231	 */
232	atomic_t reservations;
233
234	/*
235	 * Incremented while holding the spinlock *lock* by a task checking if
236	 * it can perform a nocow write (incremented if the value for the *ro*
237	 * field is 0). Decremented by such tasks once they create an ordered
238	 * extent or before that if some error happens before reaching that step.
239	 * This is to prevent races between block group relocation and nocow
240	 * writes through direct IO.
241	 */
242	atomic_t nocow_writers;
243
244	/* Lock for free space tree operations. */
245	struct mutex free_space_lock;
246
247	/*
248	 * Number of extents in this block group used for swap files.
249	 * All accesses protected by the spinlock 'lock'.
250	 */
251	int swap_extents;
252
253	/*
254	 * Allocation offset for the block group to implement sequential
255	 * allocation. This is used only on a zoned filesystem.
256	 */
257	u64 alloc_offset;
258	u64 zone_unusable;
259	u64 zone_capacity;
260	u64 meta_write_pointer;
261	struct btrfs_chunk_map *physical_map;
262	struct list_head active_bg_list;
263	struct work_struct zone_finish_work;
264	struct extent_buffer *last_eb;
265	enum btrfs_block_group_size_class size_class;
266};
267
268static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
269{
270	return (block_group->start + block_group->length);
271}
272
273static inline bool btrfs_is_block_group_used(const struct btrfs_block_group *bg)
274{
275	lockdep_assert_held(&bg->lock);
276
277	return (bg->used > 0 || bg->reserved > 0 || bg->pinned > 0);
278}
279
280static inline bool btrfs_is_block_group_data_only(
281					struct btrfs_block_group *block_group)
282{
283	/*
284	 * In mixed mode the fragmentation is expected to be high, lowering the
285	 * efficiency, so only proper data block groups are considered.
286	 */
287	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
288	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
289}
290
291#ifdef CONFIG_BTRFS_DEBUG
292int btrfs_should_fragment_free_space(struct btrfs_block_group *block_group);
293#endif
294
295struct btrfs_block_group *btrfs_lookup_first_block_group(
296		struct btrfs_fs_info *info, u64 bytenr);
297struct btrfs_block_group *btrfs_lookup_block_group(
298		struct btrfs_fs_info *info, u64 bytenr);
299struct btrfs_block_group *btrfs_next_block_group(
300		struct btrfs_block_group *cache);
301void btrfs_get_block_group(struct btrfs_block_group *cache);
302void btrfs_put_block_group(struct btrfs_block_group *cache);
303void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
304					const u64 start);
305void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
306struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
307						  u64 bytenr);
308void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
309void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
310void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
311				           u64 num_bytes);
312int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
313struct btrfs_caching_control *btrfs_get_caching_control(
314		struct btrfs_block_group *cache);
315int btrfs_add_new_free_space(struct btrfs_block_group *block_group,
316			     u64 start, u64 end, u64 *total_added_ret);
317struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
318				struct btrfs_fs_info *fs_info,
319				const u64 chunk_offset);
320int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
321			     struct btrfs_chunk_map *map);
322void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
323void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
324void btrfs_reclaim_bgs_work(struct work_struct *work);
325void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
326void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
327int btrfs_read_block_groups(struct btrfs_fs_info *info);
328struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
329						 u64 type,
330						 u64 chunk_offset, u64 size);
331void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
332int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
333			     bool do_chunk_alloc);
334void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
335int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
336int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
337int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
338int btrfs_update_block_group(struct btrfs_trans_handle *trans,
339			     u64 bytenr, u64 num_bytes, bool alloc);
340int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
341			     u64 ram_bytes, u64 num_bytes, int delalloc,
342			     bool force_wrong_size_class);
343void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
344			       u64 num_bytes, int delalloc);
345int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
346		      enum btrfs_chunk_alloc_enum force);
347int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
348void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
349void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
350				  bool is_item_insertion);
351u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
352void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
353int btrfs_free_block_groups(struct btrfs_fs_info *info);
354int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
355		     u64 physical, u64 **logical, int *naddrs, int *stripe_len);
356
357static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
358{
359	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
360}
361
362static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
363{
364	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
365}
366
367static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
368{
369	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
370}
371
372static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
373{
374	smp_mb();
375	return cache->cached == BTRFS_CACHE_FINISHED ||
376		cache->cached == BTRFS_CACHE_ERROR;
377}
378
379void btrfs_freeze_block_group(struct btrfs_block_group *cache);
380void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
381
382bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
383void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
384
385enum btrfs_block_group_size_class btrfs_calc_block_group_size_class(u64 size);
386int btrfs_use_block_group_size_class(struct btrfs_block_group *bg,
387				     enum btrfs_block_group_size_class size_class,
388				     bool force_wrong_size_class);
389bool btrfs_block_group_should_use_size_class(struct btrfs_block_group *bg);
390
391#endif /* BTRFS_BLOCK_GROUP_H */
392